linux-kernel - [BUG REPORT] MHI's resume from hibernate is broken

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <59c036b6-a3d6-403b-8bb0-566a17f72abc@collabora.com>
Date: Tue, 11 Feb 2025 13:15:55 +0500
From: Muhammad Usama Anjum <Usama.Anjum@...labora.com>
To: Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>,
 "David S. Miller" <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
 Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
 Johan Hovold <johan@...nel.org>, Loic Poulain <loic.poulain@...aro.org>
Cc: Usama.Anjum@...labora.com, linux-arm-msm@...r.kernel.org,
 netdev@...r.kernel.org, linux-kernel@...r.kernel.org, mhi@...ts.linux.dev,
 kernel@...labora.com
Subject: [BUG REPORT] MHI's resume from hibernate is broken

Hi,

I've been digging in the MHI code to find the reason behind broken
resume from hibernation for MHI. The same resume function is used
for both resume from suspend and resume from hibernation. The resume
from suspend works fine because at resume time the state of MHI is 
MHI_STATE_M3. On the other hand, the state is MHI_STATE_RESET when
we resume from hibernation.

It seems resume from MHI_STATE_RESET state isn't correctly supported.
The channel state is MHI_CH_STATE_ENABLED at this point. We get error
while switching channel state from MHI_CH_STATE_ENABLE to
MHI_CH_STATE_RUNNING. Hence, channel state change fails and later mhi
resume fails as well. 

I've put some debug prints to understand the issue. These may be
helpful:

[  669.032683] mhi_update_channel_state: switch to MHI_CH_STATE_TYPE_START[2] channel state not possible cuzof channel current state[1]. mhi state: [0] Return -EINVAL
[  669.032685] mhi_prepare_channel: mhi_update_channel_state to MHI_CH_STATE_TYPE_START[2] returned -22
[  669.032693] qcom_mhi_qrtr mhi0_IPCR: failed to prepare for autoqueue transfer -22

This same error has been reported on some fix patches [1] [2]. Are there
any patches which I should test? 

Is officially hibernation use case supported at all?

In my view, this path may not have gotten tested and can be fixed easily
as we need to perform, more or less the same steps which were performed
at init time. But I've not found much documentation around MHI protocol
and its state machine, how is mhi state related to mhi channels support.

PS: Just some information, my device has QCNFA765 endpoint and ath11k_pci
is being used. It doesn't seem like the problem is in the driver. The
problem is on the host side which isn't able to communicate correctly with
the endpoint. The resume from hibernation would be broken on all wifi
chips using mhi.

[1] https://lore.kernel.org/all/Z5EKrbXMTK9WBsbq@hovoldconsulting.com 
[2] https://lore.kernel.org/all/Z5ENq9EMPlNvxNOF@hovoldconsulting.com 
-- 
BR,
Muhammad Usama Anjum